Generate Meeting Minutes from Videos with Whisper, Ollama LLM, and Notion

This n8n workflow automates the process of generating meeting minutes from video files, transcribing them using the Whisper ASR model via Ollama, summarizing the transcription with an Ollama-powered LLM, and finally storing the structured minutes in Notion.

It simplifies the task of documenting video meetings, making it easy to extract key information and store it in a searchable format.

What it does

1. Monitors for New Video Files: Listens for new video files (e.g., .mp4, .mov, .webm) in a specified local directory.
2. Uploads to Google Drive: If a new video file is detected, it uploads the file to a designated folder in Google Drive.
3. Transcribes Video with Whisper (via Ollama):
   • Executes a shell command to run the Whisper ASR model (hosted via Ollama) on the video file to generate a transcription.
   • The transcription is saved to a local .txt file.
4. Summarizes Transcription with Ollama LLM:
   • Reads the generated transcription from the local file.
   • Uses an Ollama-powered Large Language Model (LLM) to summarize the meeting transcription.
   • The summary is then processed to extract key information like "Title" and "Summary".
5. Stores Minutes in Notion: Creates a new page in a specified Notion database, populating it with the generated title and summary.
6. Deletes Local Files (Optional): If the Notion entry is successfully created, it deletes the local transcription file to clean up the workspace.
7. Notifies on Discord (Optional): Sends a success notification to a Discord channel, including a link to the Notion page.

Prerequisites/Requirements

• n8n Instance: A running n8n instance with access to the local file system.
• Ollama: Ollama installed and running on the same machine as n8n, with the Whisper ASR model and a suitable LLM (e.g., llama2, mistral) pulled and available.
  • ollama pull whisper
  • ollama pull llama2 (or your preferred LLM)
• Google Drive Account: A Google Drive account and an n8n Google Drive credential configured.
• Notion Account: A Notion account with a pre-configured database for meeting minutes and an n8n Notion credential configured.
• Discord Account (Optional): A Discord server and an n8n Discord credential configured for notifications.
• Local Directory: A local directory on the n8n host where video files will be dropped and transcriptions will be temporarily stored.

Setup/Usage

1. Import the Workflow: Import the provided JSON into your n8n instance.
2. Configure Credentials:
   • Google Drive: Set up your Google Drive OAuth2 credential.
   • Notion: Set up your Notion API credential. Ensure the Notion integration has access to your meeting minutes database.
   • Discord (Optional): Set up your Discord Webhook or Bot credential.
3. Configure Nodes:
   • Local File Trigger (Node 533):
     • Specify the Directory to monitor for new video files.
     • Set File Pattern to match your video file types (e.g., *.mp4, *.mov, *.webm).
   • Google Drive (Node 58):
     • Select your Google Drive credential.
     • Specify the Folder ID where video files should be uploaded.
   • Execute Command (Node 13 - Transcribe with Whisper):
     • Update the Command to correctly call Ollama's Whisper model with the path to the input video file and the desired output transcription file path.
       • Example command: ollama run whisper "{{ $json.path }}" > "/path/to/transcriptions/{{ $json.fileNameWithoutExtension }}.txt"
       • Important: Replace /path/to/transcriptions/ with an actual directory on your n8n host.
   • Read/Write Files from Disk (Node 1233 - Read Transcription):
     • Ensure the File Path points to the transcription file generated in the previous step (e.g., /path/to/transcriptions/{{ $json.fileNameWithoutExtension }}.txt).
   • Ollama Model (Node 1159):
     • Select the Model you want to use for summarization (e.g., llama2, mistral).
   • Basic LLM Chain (Node 1123):
     • Review and adjust the Prompt to guide the LLM in summarizing and extracting the "Title" and "Summary" from the transcription.
   • Notion (Node 487):
     • Select your Notion credential.
     • Set the Database ID for your meeting minutes database.
     • Map the Title and Summary properties to the output of the LLM chain.
   • Execute Command (Node 13 - Delete Local File):
     • Ensure the Command correctly deletes the local transcription file (e.g., rm "/path/to/transcriptions/{{ $json.fileNameWithoutExtension }}.txt").
   • Discord (Node 60 - Send Discord Message):
     • Select your Discord credential.
     • Customize the Message content for the success notification.
4. Activate the Workflow: Enable the workflow to start monitoring for new video files.
5. Test: Drop a video file into the configured local directory to test the end-to-end process.